We have proposed to test the hypothesis that the quantitative analysis of between 200 and 300 constituents of human urine followed by computerized pattern recognition can be used to diagnose and quantitatively measure a large proportion of human diseases. We are particularly interested in the application of these techniques to diseases that lead to mental illness, but we have been testing the techniques also on a wide variety of other diseases. At the present time, we have indication that usable patterns are present for multiple sclerosis, muscular dystrophy, breast cancer, and various kinds of mental retardation. In addition, we have established patterns for various states of normal health associated with sex, fasting, biochemical individuality, and birth control pill intake . The greatest problems to be overcome in this research have been questions of human sampling and perfection of analytical techniques. In addition, we are interested in the general problem of improvement of human nutrition. Our work in this area is divided between the evaluation of present knowledge that may be used to improve human nutrition and an attempt to gain further basic knowledge about nutrition. This part of our work has led to several recommended changes in commonly held views about optimum human nutrition.